A C T A U N I V E R S I T A T I S L O D Z I E N S I S FOLIA LINGUISTICA 31, 1994
Barbara Lewandowska-Tomaszczyk
P R O T O T Y P E S , S T E R E O T Y P E S . A N D T Y P E S
1. INTRODUCTION
During the past two decades a considerable amount of theoreti cal and experimental research in psychology, linguistics, philo sophy and Artificial Intelligence has been devoted to the classical questions concerning the relationship between language and human conceptual system.
The present paper aims at giving a perspective on the overlaps on the one hand, and the disparities on the other, between some basic concepts utilized in such studies, namely the concepts of prototype and stereotype in contradistinction to the classical con cept of type.
2. THEORY OF PROTOTYPES
A number of researchers in different fields, starting with Wittgenstein and his family resemblance and earlier, have played a significant role in the history of challenging the basis of the classical conception of categorization. It is however Eleanor R о s с h [1973] first of all, who is credited with putting forward what has come to be called the theory of prototypes and basic level categories.
Basing on her experimental study of the colour terms in a New Guinea language, Dani, Rosch found out that in learning, memorizing, and the evaluation of colour terms, speakers use what she called prototypes, i.e. category exemplars which serve as special (cognitive) reference points not only for the speakers of one language but, as was confirmed in further research (1977), more universally.
Distances or asymmetries between these prototypical exemplars and other category members have been called prototype effects.
In his latest book George L а к o f f [1987] gives a detailed description of Rosch's development of the interpretation of her experimental results, and points to the fact that while in her earlier work Rosch tended to interpret prototype effects as the characterization of the representation of the categories in the human mind, in the late 70s she clearly abondoned this idea, as is quite evident from the following quotation:
1. To speak of a prototype at all is simply a convenient grammatical fiction, what is really referred to are judgments of prototypicality [...].
2. Prototypes do not constitute any particular processing model for categories [...].
3. Prototypes do not constitute a theory of representation for categories [...].
4. Although prototypes must be learned, they do not con- l o ^ l ] 6 any particular theory of category learning [ R o s c h 1978:
As can be seen, however, although prototypes alone cannot reveal the nature of the categorial boundaries, judgments of prototypicality give us important clues to the interpretation of the internal category structure.
2.1. BASIC LEVEL CATEGORIES
Categorial hierarchies have been shown to be organized not in a regular taxonomic model but rather into distributionally uneven levels. It is the middle levels of a hierarchy that represent the psychologically most salient basic levels e.g. in the hierarchy mammal - dog - dachshund it is the category dog which is basic, while in the hierarchy furniture - chair - kitchen chair, it will be the category chair. The basic level, being also the name of the whole category (e.g. dog), which includes all lower taxa, stands closest to the prototypical exemplar of that category both in terms of its gestalt, image-schematic, as well as typicality properties
[cf. P u 1 m a n 1983].
2.2. BACKGROUND FRAMING
Prototypes are not 'context-free' entities. They do not act in vacuum. As was first proposed by Marvin M i n s k y [1975] and Charles F i l l m o r e [1982], the interpretation of cate gories takes place in terms of larger knowledge patterns. They have been differently termed: frames with defaults - by Minsky, scenes
and frames - by Fillmore, Idealized Cognitive Models (ICMs) - with F i l l m o r e [1982] and L а к о f f [1987], or scripts, plans, and schemas - as with S c h a n k and A b e 1 s о n [1977].
In L e w a n d o w s k a-T o m a s z c z y k [1987] the no tion of Cognitive or Conceptual Domain was introduced, which re presents general knowledge of the paradigmatic categorial connec tions between domains and subdomains. Factual knowledge of a syn- tagmatic type, i.e. knowledge about states and events, as well as about their order and sequentiality, is assumed to be stored in the human meaning as global patterns or complete stereotypic chunks [cf. M i n s k y 1975]. Since the terminology connected with the patterns proliferate in the Artificial Intelligence studies [cf. S c h a n k and A b e 1 s о n 1977] and linguistics [e.g. F i l l m o r e 1977, L a n g a c k e r 1983], de B e a u g r a n d e and D r e s s i e r [1981: 90] proposed the follow ing systematization of the.terms:
FRAMES - contain commonsense knowledge about some central concepts (birthday parties, restaurant scenes),
SCHEMAS - global patterns of events and states in ordered sequence linked by time proximity and causality, arranged in progression,
PLANS - global patterns of events and states leading up to an intended GOAL - evaluated in terms of how they advance forward the planner's goal.
SCRIPTS - stabilized PLANS called up frequently to specify the roles of participants and their expected actions, differ from plans by a preestablished routine.
However, as can be seen, the proposed classification is not quite satisfactory. Frames, including, say, restaurant scenes, can not be in fact distinguished from Schemas or Scripts, both, accord ing to the definitions above, potentially able to cover the same restaurant scenes. The roles of participants and the sequential progressing would have to be specified regardless of the proposed type of patterning.
The expression background framing is used in some frameworks in regard to the background knowledge of the world which can affect the interpretation of the linguistic expression, e.g. the meaning of bird can be analysed in terms of Ornitological, Pet, Hunting, etc. background framing. Background framing can have a more static form as in e.g. ORNITOLOGICAL FRAME, or a more dynamic one, as in the HUNTING SCRIPT. The notions of PLANS and GOALS are retained
in their original sense as introduced by S c h a n k and A b e 1 s о n [1977].
From the point of view of their organization both frames and scripts are data structures which consist of:
(a) collection of concept-nodes with a characteristic fea ture specification for each concept and markers pointing to other frames/scripts,
(b) characteristic relations for those concepts.
Concepts represent either prototypical structure or extensions of a prototype. They too can be represented as a structure of nodes, connected with one another as a network system. Some nodes both in frame/script and in prototype configurations are associated with partial but "constant" knowledge while some others are filled with variable 'default' assignments, true, unless some disconfirming information is provided. Defaults are loosely attached to the nodes, so that new information can be added there, or some changes can be incorporated.
From the point of view of typicality of their structure both frames, scripts and (object) prototypes may have either prototypi cal structure or, undergoing a 'prototype effect' represent exten sions of a prototype.
Some modification of the concept such as, say toy pistol or fake gun influences the nominal concepts in terms of defeating their most essential properties to the extent that we can say we no more have to do with the extension of the prototype pistol or gun but we rather deal with an entirely new entity. Thus may be true even with such cases as social lie, which is considered by some language users as no lie at all.
An example of the FAMILY FRAME after T h o m p s o n [1976: 13] is given below:
Family Abstract
(self a Simple Family with
daddy « the male Parent mommy “ the female Parent Kids « the children) (male Parent a Person with
sex “ Male)
(female Parent a Person with sex « Female)
(parents the male Parent, the female Parent; a Couple with
male Half « the male Parent, female Half m the female Parent) (male Children some *a Person with
sex ■ Male’)
(female Children some 'a Person with
sex ■ Female’)
(children the male Children the female Children) (generation a Generation with
pairs “ each in the children, a Person)
Fig. 1
The type of knowledge presented above embraces the prototypical in stance of the FAMILY FRAME, where such terms as brother or son are referred to in further specification of family relations.
Stimulated in a perceptual or cognitive act, a matching process starts after a frame (script, etc.) is activated.
Particular token knowledge is matched then with the general frame and a concrete family (or a member of the family) can be defined in terms of the frame content. In this context the default 'children' can be overridden by contingent facts, if there is only one son or daughter in the family.
There are cases of complex frame/script configurations even if they underlie the meaning of a single lexeme in a language. One such case of frame/script stacking can be found in what I called the Superordinate Frame of ADVERTISING [ L e w a n d o w s k a 1984]. This main frame can instantiate subframes of different sort with a set of 'terminal' nodes, which, in turn, can evoke other frames, or may serve as connection pointers between ADVERTISING script/frame and other frames.
As global patterns of knowledge can incorporate PLAN and GOAL schemes, the Super F of advertisements reflects the persuasive character of this concept (Fig. 2):
Fig. 2. Superframe of advertisement
The 'labirynthian' structure of the conceptual meaning of advertising is presented in the prototypical form in Fig. 2. However, there are numerous extensions of the prototype, one of them for instance in volves TRADE frame. As a less typical variant, the concept of trading exhibits the following processual script as in the case of the trading of information, software, services etc., where the seller does not get rid of the 'object' sold.
A has Y, В has X --- ►
<---__---A has X, В has Y' and X Fig. 3
There are certain problems with frame representations connected with the criteria for the groupings of semantic information as found in frames. As M e t z i n g [1981: 340] notices the informa tion may be either independent of the speaker (as in the case of stereotyped events), or it can be speaker-dependent (as when based on internal inference processes). If the latter is the case it may
В has X try to make instantiate the desired X frame: method: suggested instantiate trade frame: A has Y, В has X ■ A has X, В has Y -A has Y ADDRESSEE (A)
I
trade (+) values --- ---1 ---ITEM (X) for YI
appear that the speaker is able to propose such 'groupings of semantic information' which defeat conventional patterns found in the language of his/her audiences.
Knowledge types as described above (general vs particular) has been taken by C o p e l a n d [1984: 232] as a distinction over which knowledge states can be defined in the interactional context. The figure below (Fig. 3) is modified from Copeland:
KNOWLEDGE STATES KNOWLEDGE TYPES present activated not activated not present -rGeneral and Particular (-deactivated -unactivated 1-General only гGeneral only General and Particular [-General and Particular General only Particular only General and Particular (Given) (Some) (Recoverable-Some) (Recoverable) (Computable) (New) (Novel) Fig. 4
The speaker's knowledge of the world can be either shared (present) or not shared (not present) by his/her audience. If it is present, it can be either activated or not at a given moment. If it is not activated, it may mean that it was activated previously in the given discourse context, but it no longer is i.e. it is deactiva ted, or it may be as yet unactivated. Both the syntactic and proso dic behaviour of the speakers as well as the audience have been found to provide evidence for these knowledge states.
3. PUTNAM'S REALISM AND STEREOTYPES
George L а к o f f [1987: 116] in his recent book acknow ledges the contributions of Marvin Minsky and Hilary Putnam to the explanation of prototype effect in terms akin to cognitive models accounts. The aim of this section is to show that it may be par tially misleading to put Putnam in the same compartment as Minsky or Fillmore, and to elucidate the approach Putnam introduced, which is based on the idea of stereotype. I will also try to illuminate basic differences between the philosophical positions held by Put nam and those taken by the majority of cognitivists. This com parison should be especially revealing in the light of what seem to be diametrically opposite views on Putnam's position expounded by J о h n s o n-L a i r d [1983] and L а, к о f f [1987]. John- son-Laird considers Putnam the strongest realist possible, while L а к о f f [1987: 229] proposes that Putnam "has provided a de vastating logical critique of the view of meaning and reference in what he calls metaphysical realism, which is a generalized version of what we have called objectivist semantics".
As will be seen below [cf. L e w a n d o w s k a-T о m a- s z с z у к 1987], Putnam's critique of model-theoretic semantics [1981] does not necessarily imply his unambiguously anti-realist position, as L а к о f f would probably like it [1987: 229]. Putnam argues first of all against the use of formal systems and the interpretation of truth connected with them as the models of meaning in natural language. This position however does not presup pose or entail a univocal conceptualist stand.
Since the time of F r e g e's distinction [1892] between 'Sinn' and 'Bedeutung' the most frequently raised issue in philo sophy of language and linguistics concerns the status of sense and reference, later intension and extfension.
Two major trends, Realism and Mentalism, describe them in different ways and assign them different origin and existence. Realism treats intensions as objective 'givens', having independent reality outside the human language user's mind. All psychologically based theories on the other hand, situate them in the human mind and consider meanings as cognitive entities.
But even the most radical realists cannot deny that the human mind does play a role in capturing the sense of natu ral language and that some kind of cognitive entities, be they in tensions or not, appear as a result of experiencing reality by
the human being. The entities Hilary P u t n a m [1975] is trying to describe, are distinct from intensions, which are supposed to exist independently of the human mind, and are referred to by him as stereotypes. It is stereotypes, as Putnam is trying to argue, and not intensions, that exist in the human language user's head and are the basis of meaningful communication, while intensions may or may not be subject to human penetration and cognition.
Stereotypes may be true, i.e. they may correspond to reality, or they may be false, originating in the human lack of knowledge or false beliefs. This, however, does not block successful communi cation, as people in fact rely on judgements and opinions of better informed members of a community, according to what P u t n a m calls 'a division of linguistic labour' [1975: 227].
Stereotypes represent theories language users have about referents of the words used. The stereotype then associated with e.g. the word lemon embraces the following features: lemon is a natural kind word and refers to fruit with yellow peel, tart taste, etc. These are the core properties without which even the approximate sense of the word could not be conveyed.
The same stereotyp« can be associated with different terms, e.g. the same characteristics that are assigned to, say, aluminium, may fit molybdenum (a light metal which makes durable pots and pans and does not rust). For that reason, as Putnam is arguing, part of the meaning of a term must be connected with its extension represented either as, say, its chemical structure (water - H20), or by a sample (for colours), or weight perhaps, the atomic weight (dif ferent for aluminium and for molybdenum) in the case of a metal. Such information, frequently referred to as 'encyclopedic informa tion' cannot, in actuality, be distinguished from purely linguistic data.
Since in Putnam's approach the extension of a term is not determined relative to the psychological state of the language user and is assumed to be superposed by the intension (e.g. all mem bers of the class FRUIT must be covered by the intension of the word fruit), then the only course open to Putnam is to propose that the psychological state of the language user cannot determine the intension of the terms either, so the intension of the term is an objectively given really existing abstract entity. Such an ap proach to meaning closely resembles classical versions of (Plato- nist) realism [cf. K a t z 1981]. What is affected on the other hand by the psychological state of the language user is the stereo
type of the term which may or may not (this is supposedly not immediately evident to the human being) correspond to the intension of the term.
Putnam gives a fairly controversial example of two hypo thetical individuals, an Earthian O s c a r a n d a Twin Earthian Oscar2- Suppose the two most frequently used and perceived liguidä on both planets have all the perceptual and functional properties of water, the extension of which however is H20 on Earth and XYZ on Twin Earth. In, say, 1750, however, this fact was not known to anybody (including Oscar^ and Oscar2) either on Earth or on Twin Earth. To prove that the extension is not determined by psychologi cal factors, P u t n a m [1975: 224] argues:
Oscar^ and Oscar2 understood (emphasis mine) the term "water" differently in 1750 although they were in the same psychological state (emphasis mine), and although, given the state of science at the time, it would have taken their scientific communities about; fifty years to discover that they understood the term "water" differently.
Although one could tentatively agree with the conclusion that the extension of the term water is not a function of the psy chological state of the speaker, one can raise objections against Putnam's claim that the two individuals understood the terms dif ferently although they were in the same psychological state of mind It seems hardly controversial nowadays to assume that understanding is, among other things, a function of the psychological state of the mind, even if we accepted that actual extensions of terms might not be. The core part of the problem seems to be the question as to whether intensions and extensions should be best considered onto logical or epistemological entities. Understanding, however, i.e. grasping these entities, is an epistemological matter determined by psychological states of the human being.
Developing his argument further, P u t n a m [1975: 226] dis cusses the meaning of the two words: elm and beech. He argues that although he cannot tell one tree from the other (i.e. he cannot point to the denotata of the extensions), the extensions in fact are different. "Is it really credible", he asks, "that this difference in extensions is brought about by some difference in our concepts? My concept of an elm tree is exactly the same as my con cept of a beech tree (I blush to confess). (This shows that the identifications of meaning 'in the sense of intension' with concept cannot be correct, by the way)".
Putnam seems to completely disregard this very important piece of information that he knows that the trees are different, no matter what the exact differences are, so, although his concepts of a beech tree and an elm tree may be similar, they are not identical. He would never use these labels as synonyms! His concepts of these objects are incomplete, while the verification of their identity if left to experts, according to Putnam's own principle of the di vision of linguistic labour (p. 228), which is possibly a good can didate for a sociolinguistic universal. Putnam argues in this con nection that
Whenever a term is subject to the division of linguistic labor, the 'average' speaker acquires it does not accuire anything that fixes its extension. In particular, his individual psycho logical state certainly does not fix its extension, it is only the sociolinguistic state of the collective linguistic body to which the speaker belongs that fixes the extension, (p. 229). The speaker may by capable of pointing to these instances of the category to the extent to which he has developed a given concept. This, however, means simply that both one speaker's con cept (different from Putnam's intension) of the term as well as his ability to identify the category members (corresponding to exten sions) will not be identical with those of an expert. However, this should not exclude the meanings used by 'average' speakers from the scope of linguistic interests.
As has been mentioned above, people have stereotypes which serve as a basis of member identification in the given class. Stereotypes are assumed to have criteria - necessary, or pro babilistic necessary conditions providing clues for recognizing natural kind members. By postulating stereotype criteria Putnam departs from a necessary and sufficient feature model of mean ing in the epistemological part of his theory. Whether be would like to retain the oriterial model in his ontological part con cerning 'true' intensions of terms is a different matter. Putnam accepts is as possible that objects may have some 'hidden structure'
(possibly a genetic code), which might determine class membership not only in the socially accepted reality but in all 'possible worlds'. Such essentialism would justify some plausible counter- factual suppositions about natural kind terms (e.g. Water could have all been vapour - TRUE) and discard others (Water could have all been XYZ - FALSE). On the other hand, Putnam seems hesitant about the 'essential' structure - in cases where the structure may
not be uniform and it is the superficial properties that play the decisive role in concept formation and recognition.
Stereotype criteria are assumed to be culture and topic dependent, but familiarity with these features is expected, even required in the linguistic community in order for a com munity member to be considered a successful acquirer of the stereo type (e.g. the feature striped for 'tigers'). It is in this sense that one can say that stereotypes are obligatory for community mem bers, as Putnam puts it "in the same sense of 'obligatory' in which it is obligatory to indicate whether one is speaking of lions in the singular or lions in the plural when one speaks of lions in En glish" (p. 251).
The consequence of the 'objective truth' position that Putnam assumes is still more evident when he talks about "wrong" or "widly inaccurate" stereotypes (pp. 249-250). As an example Putnam gives the feature yellow for gold, whereas chemically pure gold is known to be white. Quite evident here is the lack of a contextual frame of reference that should be incorporated in the analysis of stereo types. And even though Putnam provides an explanation of why the feature yellow is present in the "wrong" stereotype ("the gold we see in jewelry is typically yellow due to the presence of copper"), he does not propose any more constrained apparatus to include a possibility of contextual variation in his model of meaning.
Some features of stereotypes seem more important than others, e.g. ANIMAL - for tiger, DAY OF THE WEEK - for Tuesday, PERIOD OF TIME - for hour, etc. They are what Q u i n e [1951] would call more central or unrevisible category-indicators. Due to their signi ficant function Putnam gives them a more significant position in his system assigning them the status of semantic markers.
Summing it up, the meaning of a word for Putnam is "a finite se quence, or 'vector', whose components should certainly include the following (it might be desirable to have other types of components as well):
(1) the syntactic markers that apply to the word, e.g. 'noun', (2) the semantic markers that apply to the word, e.g. 'animal', 'period of time',
(3) a description of the additional features of the stereotype, if any,
(4) a description of the extension.(p. 269).
A partial description of the meaning of, say 'water' that Putnam gives, is as follows:
SYNTACTIC MARKERS SEMANTIC MARKERS STEREOTYPE EXTENSION mase noun concrete natural kind liquid colorless transparent tasteless thirst-quenching etc. H20 (give or take im purities)
What emerges from Hilary Putnam's theory of word meaning, and is further confirmed by Saul К r i p к e [1972] is not in fact an argument against the criterial features model as seems to be sug gested by S m i t h and M e d i n [1981], but first of all a necessity of a proper distinction between metaphysical and epistemo- logical possibility and necessity [cf. R e y 1983]. An interest ing thing is that a similar conclusion can also be drawn in re ference to the description of meaning based on prototypes and exem plars [ R o s c h 1975, etc.]. These approaches to meaning seem first of all to exploit epistemological possibilities of concept formation and recognition, leaving aside metaphysical possibility and necessity altogether.
R e y [1983: 241] proposes the following list of the main func tions that concepts perform:
I. STABILITY FUNCTIONS:
(a) intrapersonal: the basis for conceptual competence and for comparisons of cognitive states within a given agent,
(b) interpersonal: the basis for comparisons of cognitive states across agents.
II. LINGUISTIC FUNCTION: the meaning of open class linguistic items, whereby they enjoy relations of translation, synonymy, anto- nymy, and semantic implication.
III. METAPHYSICAL FUNCTIONS:
(a) metaphysically taxonomic: that by virtue of which things are the kinds of things they are (by virtue of which they are cor rectly classified).
(b) metaphysically modal: the basis for claims of counter- factuality, possibility, and necessity.
IV. EPISTEMOLOGICAL FUNCTIONS:
(a) epistemologically taxonomic: the means by which an agent categorizes things, decides whether or not something is of a certain kind,
(b) epistemologically combinatorial: the means by which an agent categorizes things into complex kinds,
(c) epistemologically modal: the basis for claims to 'a priori knowledge', or knowledge justifiable 'completely independently of any experience'.
The linguistic function is in fact an expression of the stability functions. They provide a basis for constructing a typology of different content of different cognitive states and in this way make it possible to characterize the relations of in ference, linguistic equivalence, implication, etc.
The core of the controversy which is replicated in dif ferent approaches to meaning in natural language is the is sue concerning the distinction between the epistemological and me taphysical functions. In simple terms, the controversy refers to a division between, as Rey puts it (p. 243): "issue surrounding how the world is (what exists, what is true) and issues surrounding how we know, believe, infer, how the world is". There are cases in which this distinction is not too sharp, nevertheless it is psychologi cally real. People have no doubt that the questions whether this me tal is gold and whether someone believes (knows, infers etc.) that this metal is gold are basically different. Similarly the questions whether this metal is gold and how I know this metal is gold, being based ОП ПОП- -identical premises, may yield different answers. There may exist defining conditions for gold even though the most competent langua ge user would not know them. It is then metaphysically possible that gold could have some defining conditions, not known to us, the con ditions which may be different from those we seem to know. But if it is true that its atomic number is 79, then it is necessarily true. So metaphysical possibility and necessity, as Kripke and Putnam argue, have nothing to do with the human language user. They con- cern the world as it really is.
Epistemological possibility, on the other hand, may be read as: 'to the best of the language user's knowledge', 'compatible with a system of his beliefs' etc. What is important is the fact that this concept does n o t exclude any other possible state of af fairs if such a situation occurs, in other words, it is assumed to be contingent. Metaphysical necessity and epistemological possibi lity can be made compatible. To use R e y's [1983: 254] wording: "Given how little I know about chemistry, I am prepared to discover that gold has atomic number 89, even though, if it in fact has atomic number 79, then necessarily it does".
For that reason linguistic models of meaning should and do in fact concentrate on the analysis of meaning in the epistemological sense. It should be emphasized here that even such semantic theories which are putatively 'free from psychologism' (e.g. truth- -conditional semantics) are also based on cognitive principles in this sense.
The significance that both Putnam and Kripke associate with the above distinctions shows quite clearly that they do not argue against the classical view of necessary and sufficient conditions. On the contrary, they try to show that essentialism is what con stitutes the world, although the true structure of the matter may be in fact beyond the human cognitive capabilities. For that reason, although their theories can be taken as an attack on definability of concepts, they are not aimed against the defining conditions model. Such a view, as Rey argues (p. 252: ft 15), seems to have been anticipated even by Wittgenstein, e.g. in his discussion of 'symptoms' and 'criteria' of angina which vary over time
[ W i t t g e n s t e i n 1965: 25-26].
Language, as used by competent language speakers, is based, among others, on words and concepts understood in the epistemolo gical sense, relative to the language users' sets of beliefs and convictions. It is most probable that meanings of language units, as used by people, approach metaphysical truths in these regions where such truths do exist and where they can be penetrated. At some other points, however, where the subject of our interest are either terms vague 'in essence' or phenomena familiar to us only through 'outer' features, we are completely at a loss as to what 'real' criterial features of such entities may be. And although I am far from denying the possibility of existence of 'real' metaphysically necessary properties of concepts, I think that it may be the case that competent language users will never have access to the core structure of some concepts. This, however, does not and will not make it impossible for them to enter verbal interactions and to communicate successfully.
The inaccessibility of 'real essence' of concepts may also lead to a total scepticism as to the possibility of defining meaning. Such a position is known primarily from classical structu ralism [ B l o o m f i e l d 1933], which eliminates semantics from any serious linguistic activity. On the other hand, such at titudes like Wittgenstein's quietism expressing scepticism against definitional feature analyses, has led to a more constructive
ap-proach to meaning based on his 'family resemblance' model of categorization, and Rosch's theory of prototypes and basic levels.
3.1. CONCEPT MAXIMUM AND CONCEPT MINIMUM
In his classical paper on meaning, P u t n a m [1975: 14] states that "not all criteria used by the linguistic community as a collective body are included in the stereotype, and in some cases the stereotype may be quite weak". W i e r z b i c k a , in her recent book Conceptual Analysis and Lexicography [1985] takes issue with Putnam. She argues instead that [1985: 215] "all the criteria used by the linguistic community as a collective body are included in the stereotype, but not all members of the linguistic community know all the stereotypes. Which stereotypes they know, and whether they know them completely, depends on their individual life ex periences" .
In this way Wierzbicka eliminates Putnam's concept of weak stereotypes. Instead, Wierzbicka suggests that this dichotomy could be possibly applied to language as a whole to point out differences between the rich stereotypes of say, dog and horse in English, and the weaker stereotypes of tiger or lion in this commu nity.
To account for the differences found in individual speakers on the other hand, Wierzbicka posits a distinction between what she calls the concept maximum versus the concept minimum.
By the concept maximum W i e r z b i c k a understands "a ma ximum of ideas which people in a speech community see as a shared stereotype" [1985: 215]. For English speakers the concepts such as apple or potato should be instances of the concept maximum. The examples of the concept minimum on the other hand seem more in dividually determined, but such instances as papaya or even pineap ple may be good candidates in the Polish community for the concept minimum, i.e. "meanings one knows incompletely (but sufficiently to use them)" [ W i e r z b i c k a 1985: 218].
4. TYPES
E v e n s et al. [1980] report the psychological research of R u m e l h a r t , L i n d s a y , and N o r m a n [1972], who pro pose a distinction between primary and secondary concepts. They
con-sider primary concepts to correspond roughly to lexical entries representing types, while secondary concepts - tokens, are to represent individual objects. Secondary concepts are linked with the primary node by the relation called ISA e.g.:
(1) A robin ISA bird i.e. A robin (A) IS-A-MEMBER-OF a class (set) of birds (B).
In this sense then type functions as a superordinate genus in the taxonomic definition associated with the item specific differentiae.
The distinction between the terms type and token however is not an innovation of the last few years. It was first introduced in the works of C. S. P e i г с e, and as Lyons notices [1977: 13], it originally referred to the types and their actual instantiations. Only then was its sense extended (especially in psychological research) to cover also hyponymic relations.
The concept of type in lexical semantics strictly relates to the problems of generic reference. Some sentences express ing generic propositions refer not to particular exemplars of a category but to the whole class embracing its members i.e. to the type, e.g.:
(2) The puma lives in mountain regions of America. Within the framework of predicate calculus representation, the NP (the puma) in (2) would be formalized as involving universal quantification, so it would be interpreted as "for all values of x etc." In this form there would be no distinction between the semantic formula for (2) and that for (3) :
(3) All pumas live in mountain regions of America. This synony mity is certainly not intended by the speaker of (2) and, as L y o n s notices [1977: 195] there is also a difference between the truth-conditions of (2) and (3), if one wants to stick to truth-conditional semantics: living in the mountain regions of America is not a necessary characteristics of all lions. I want to suggest then that what generic propositions of this type really ex press are actually propositions concerning prototypical exemplars of the category. Under this interpretation, the original sense of the type/token distinction, (i.e. the full instantiation of the type by a token), gets weaker. In other words, the interpretation of the meaning of the type concept must be sensitive to contingent facts. At the same time, and from a different point of view, the interpretation gets stronger, as the properties associated with the type [cf. ex. (2.)] are not merely incidental properties, but express their genuine typicality [cf. L y o n s 1977]. The
in-terpretation of the concept of type then merges at this point with the notion of prototype.
5. CONCLUSIONS
The concept of a prototypical category member partly overlaps with that of a classical type in that the type is shown to possess the highest number of properties common with the prototype. On the other hand however, the type is the notion superordinate for all - prototypical and peripheral - category members by the hyponymic relation between them, although, as has been shown above, such a relation does not necesarily entail the reference to all tokens (exemplars) of the category.
Each lexical item (not only prototypical members) is assumed to be associated with a certain stereotype. The notion of stereotype then, unlike that of a prototype or a frame, does not reflect any structure or organization of the overall scheme of conceptual/se mantic entities. Individual differences within the depth and range of the lexical senses are reflected in the strength of the stereo type with Putnam, and the distinction between the concept maximum and the concept minimum with Wierzbicka.
As to the philosophical implications entailed by a given position with respect to the linguistic meaning, there is a spectrum of different metaphysical perspectives possible, starting from the strongest externalist realism, through weaker internal realism, up to different shades of radical conceptualism. Each of these stances may imply different interpretations of the theory of truth and rationality and can be associated with varying stands towards the issue of relativism.
One of the test issues here may be the solutions concerning the fixing of reference. Conceptualists tend to argue that the referen ce of terms is fixed by the (individualistic or collective) men tal state of the language user, while realists like P u t n a m
[1981] try to show that it is the extension of a term (i.e. the set of objects in the socially accepted reality or in another possible world), in other words 'the substance itself' that fixes the re
ference of the term.
The linguistic sign, however, that refers to the 'substance' does not correspond to it independently of who uses it and how. Linguistic signs are entities internal to the conceptual schemes of their users. Such a philosophical perspective reconciles the
matter-of-fact realism of our days with the conceptualism relative to the linguistic community and to the individual.
REFERENCES
B e a u g r a n d e de R. A. and D r e s s i e r W.t 1981, Introduction to Text Linguistics. London: Longman.
B l o o m f i e l d , L., 1933, Language, New York: Holt.
В о r b ë, T. (ed.), 1984, Semiotics Unfolding, The Hague: Mouton.
С о p e 1 a n d, J. E., 1984, On the Cognitive Status of the Initial and Node in Potential and in Nonce Discourse, [in:] H a 1 1, R. A. (ed.), The Eleventh Lacus Forum 1984-1985, 229-242.
D a v i d s o n , D. and H a r r m a n G. (eds), 1972, Semantics of Natural Language, Dordrecht: Reidel.
E i k m e y e r , H-J. and R i e s e r H. (eds), 1981, Words, Worlds, and
Contexts, Berlin: Walter de Gruyter.
E v e n s , M., L i t o w i t z , B., M a r k o w i t z , J., S m i t h, R., and W e r n e r 0. (1980), Lexical-semantic relations. Edmonton, Canada: Lin guistic Research, Inc.
F i l l m o r e , Ch., 1982, Toward a Descriptive Framework for Spatial Deixis,
[in:] J a r V e 1 1 a, R. J. and K l e i n W. (eds), 31-58.
F r e g e , G., 1892, Uber Sinn und Bedeutung, "Zeitschr. f. Philosophie und Phi losoph, Kritik", 100: 25-50.
G u n d e r s o n , G. (ed.), Language, mind and knowledge. Minnesota: University
S
J a r v e 1 1 a, R. J. and of Minnesota Press. K l e i n W. (eds), 1982, Speech, place and ac- Л ' “ tion. New York: John Wiley and Sons.J о h n s о n-L a i r d, P. N., 1983, Mental models, Cambridge, Mass.: Harvard University Press.
К r i p к e, S., 1972, Naming and necessity, [in:] D. D a v i d s o n , and G. H a r m a n (eds), 253-356.
L а к о f f, G., 1987, Women, fire, and dangerous things, Chicago and London: The University of Chicago Press.
L a n g а с к e r, R. W., 1983, Foundations of cognitive grantnar, vols I and II, series A, No. 99, 100. Trier: LAUT.
L e w a n d o w s k a , В., 1984, Frames of Advertisements, [in:] Т. В о r b ë, (ed.), 1183-1191.
L e w a n d o w s k a-T o m a s z c z y k , В., 1982, Stereotypes and Creativity,
Paper delivered at the International Conference of Contrastive Studies, Błażejewko, Poland.
L e w a n d o w s k a-T o m a s z c z y k , B., 1987, Conceptual analysis, lin guistic meaning, and verbal interaction, Łódź: Acta Universitatis Lodzien- sls.
L y o n s , J., 1977, Semantics, vols. I and II. Cambridge: University Press. M e t z i n g, D., 1981, Frame Representation and Lexical Semantics, [in:]
E i k m e y e r , H-J. and H. R i e s e r (eds), 320-342.
M i n s к у, М., 1975, Frame System Theory, [in:] Schank, R. and B. L. Nash- -Webber (eds), 104-116.
P u l m a n , S. G., 1983, Word meaning and belief, London and Canberra: Croom Helm.
P u t n a m , H., 1975, The Meaning of 'Meaning', [in:] G. G u n d e r s o n , (ed.), 215-271.
P u t n a m , H.,1981, Reason, Truth and History, Cambridge: University Press. Q u i n e , W. V., 1951-1953, Two Dogmas of Empiricism; [in:] W. V. Q u i n e ,
20/46.
Q u i n e, W. V., 1953, From a Logical Point of View, Cambridge, Mass.: Harvard University Press.
R e y, G., 1983, Concepts and stereotypes, "Cognition" 15, 237-262. R о s с h, E., 1973, Natural categories, "Cognitive Psychology" 4, 328-350. R о s с h, E., 1975, Cognitive Representation of Semantic Categories, "Journal
of Experimental Psychology: General" 104, 192-233.
R о s с h, E., 1977, Human categorization, [in:] N. W a r r e n (ed.), 11-72. R о s с h, E., 1978, Principles of categorization, [in:] E. R о s с h, and
B. B. L 1 о у d (eds), 27-48.
R о s с h, E., and L l o y d В. B., (eds), 1978, Cognition and Categoriza tion, Hillsdale, N. J.: Lawrence Erlbaum Associates.
R u m e l h a r t , D., L i n d s a y , P. and N o r m a n D., 1972, Л Process Model for Long Term Memory, [in:] E. T u 1 v i n g, and W. D о n a 1 d- s о n (eds), 198-248.
S c h a n k , R., and A b e 1 s о n, R., 1977, Scripts, Plans, Goals and Under standing, Hillsdale, N. J.: Erlbaum.
S c h a n k , R., and N a s h-W e b b e r B. L. (eds), 1975, Theoretical Is sues in Natural Language Processing, Cambridge, Mass.: MIT.
S m i t h , E. E., and M e d i n, D. L., 1981, Categories and Concepts, Cam bridge, Mass.: Harvard University Press.
T h o m p s o n , H., 1976, A Frame-semantics Approach to Kinship. Linguistic Dept., UCLA, mimeo.
T u 1 v i n g, E., and D o n a l d s o n W. (eds), 1972, Organization of Memory, New York: Academic Press.
W a r r e n, N. (ed.), 1977, Advances in Cross-cultural Psychology. London: Aca
W i e r z b i c k a , A., 1985, Conceptual Analysis and Lexicography. Ann Arbor: Karoma Publishers.
W i t t g e n s t e i n , L., 1965, [1922] Tractatus logico-philosophicus. Lon don: Routlege and Kegan.
W i t t g e n s t e i n , L., 1953, Philosophical Investigations, New York: Mac millan.
Barbara Lewandowska-Tomaszczyk
PROTOTYPY, STEREOTYPY I TYPY
Artykuł ma na celu przeanalizowanie trzech pojąć używanych we współczesnym opisie semantycznym języka naturalnego: prototyp, stereotyp i typ. Jak wykazano w pracy, mimo częściowego pokrywania się zakresów pojęciowych tych terminów, róż nią się one zarówno w miejscu i funkcji, jaką przypisują im różne kierunki w opisie semantyki leksykalnej języka, jak i w założeniach filozoficznych, które się z nimi wiążą. Termin prototyp - to wynik analizy pojęciowej w ujęciu kogni tywnym, pojęcie stereotypu w teorii Putnama odzwierciedla jego pozycję realizmu wewnętrznego, zaś termin typ związany jest ze strukturalno-taksonomicznym uję ciem języka oraz z formalnym opisem jego semantyki.
